"50 AI Agents Running My Company" Is a Lie. Here's How I Build It | Gumloop, Max Brodeur-Urbas
In a Nutshell
Max Brodeur-Urbas, founder of Gumloop—an automation platform running 4 million workflows daily for companies like Instacart and Shopify—debunks hype around "50 AI agents running my company" as unreliable slot-machine slop peddled by course sellers, emphasizing that real value comes from deeply understanding a business problem before applying AI to automate repetitive tasks. After quitting Microsoft, surviving a 5-year US deportation ban, and failing 10 ideas, he pivoted from AutoGPT's unreliability to build Gumloop for non-technical users, hitting revenue in YC by focusing obsessively on users, proving ideas wrong quickly, and hiring excited ex-customers. Key advice: automate only what you understand, ignore moat paranoia, build with blind confidence, and prioritize fun culture over forced hustle.
These notes were generated by AI and may contain inaccuracies.
There's a category of entrepreneurs who think they can build a business with a single click, effortless, because they bought a course with the recipe to making a million dollars a year as a side hustle. There's an anti-pattern with AI where people claim '50 AI agents running my company' or 'a suite of AI that tells me what to do every day.' That's the wrong approach—it's like making a slot machine. Slop, not slot. Everyone on Twitter claims 'I automated everything and work one hour a week and make $10 million on the weekend with my SaaS app.' Most of that is marketing; they're lying. You can sell hope easily, selling a vision of skipping hard work and shortcutting to value, which will never happen. The course sellers make money printing it.
My name is Max, founder and CEO of Gumloop. Gumloop is an automation platform automating about 4 million workflows every day across huge businesses like Instacart, Shopify, DoorDash, Gusto. We're a team of 15, scaling quickly. It lets people who understand a business problem—like marketers, salespeople, or ops—automate their own work instead of speccing it out for engineering.
I went to McGill in Montreal, studied software engineering, did well in school, studied all the time, aimed for best grades. Goal throughout college: get a good job in big tech. Realized quickly I hated it. Common advice is go to big tech to learn then start your own thing—that's cope. Many get golden handcuffs and never leave. I don't think I've used anything learned at Microsoft in my startup. Only value: default respect from the logo on resume, validating competence. Most things I do now are the exact opposite of how Microsoft works.
What you can do at 21, 22, 23 with no responsibilities are years you can't get back. Spending them 9-to-5 fixing tickets wastes them. 'If we're going to do it, why not now? Figure out reliable path later.' That was right—just threw ourselves into the ether.
I got deported from the United States and banned for 5 years—nothing illegal. Quit Microsoft, moved to Vancouver, planned to build stuff in bedroom for a year. Visiting roommates in Seattle one weekend, border turned me around after questions (suspicious of staying longer than 2 days), resulted in 5-year ban. That moment: realized I had to build a company, no fallback. Terrified, drove back in shock, then focused, worked 6 months as hard as possible.
Tried everything: video game moderation software in VR, general trust and safety tooling, bot detection for web traffic, anti-scam platform. Made MVP weekly, tried to sell, saw market interest. Learned quickly bad ideas. Got used to proving myself wrong—in startups, chase proving yourself wrong to save weeks/months. Wasted 3 months early building for months hoping validation; opposite of right approach. Hunt for strong reason why it won't work. If none, pursue. Build as much as possible for info. Wouldn't have built Gumloop without failing 10 times before.
Talking to users is a privilege to earn—early, beg people to try product. Put users forefront, even if they say product sucks or doesn't solve problem—that's most valuable feedback. First day on current Gumloop version: stayed at office till midnight, so excited couldn't stop. That compounds; find thing getting you to midnight daily, momentum builds.
Remember AutoGPT? Popular open-source agent framework, took Twitter by storm—first AI doing something agentically. Saw on Twitter, tried, amazing demo. Joined Discord (growing exponentially), saw questions: what is GitHub? Terminal? Install locally? Dependency? Built AgentHub (first version) as nice UI to solve that, to learn frontend. Didn't think special.
Sent link in Discord for setup help. After days: 'This could be GitHub for agents.' Crumbled when realized agents unreliable. Aha: platform people wanted, but frustrated by unreliability. Gave reliability, predictability. Use cases simple, built framework to automate steps sequentially—grew into automation platform. Initially for developers (open-source), but 80% non-technical audience (business admins, ops, HR) went crazy. Realized: build approachable, fun, non-frustrating for them—motivation for simplified AutoGPT.
Got into YC 5 months before batch; product free then. First week batch: turned on pricing at $20/month (couldn't imagine more than ChatGPT). First paying customer Kai paid $20—freaked out, greatest moment, Stripe notification; still user.
Stuck in Canada during YC: pressure to build amazing without distractions, coded fast in small Vancouver studio. Learned staying focused, skipping networking/parties powerful—builders not at events. Kept attitude: don't go to events; co-founder never does. Network emerges from focus and user talks. Fundraising: build exceptional, investors come to you. Show you'll succeed without them, they waitlist you. Not complicated—just build great. Network not made at cocktail parties.
Anti-pattern: '50 AI agents running my company'—wrong, slot machine. Know what to use AI for: keep human touch in important parts, automate repetitive. Best users AI-enabled, not replacing jobs. Slippery slope. Twitter claims of automating everything, 1 hour/week, $10M SaaS—marketing lies, course bros shilling dream. 'Make $30k weekend copying workflow' never works—selling productivity vision without novel value.
Value from applying AI to what you understand deeply. No magic $30k weekend solution shared on Twitter. Productive people: take what you understand, apply AI, scale. Hype bubbles (crypto, NFTs, AI) exploit vulnerable selling hope with fictitious content.
Wantrepreneurs think single-click business effortless via course—selling skipping hard work, won't work, but sellers print money.
Only automate what I understand. Automating without understanding creates slot machine, uncertainty. Using AI to code without knowing code = making malware. Vibe coding limited. Same for workflows—if you couldn't do it yourself, result poor. Use AI to speed up what I understand, learn more, grow—not shortcut understanding or replace skill set.
Possible last generation of great engineers needed understanding before AI acceleration; now can skip. Smaller community will use AI as learning tool, grasp fundamentals. Slippery slope not understanding why things work. Bigger split: exceptional (pause, understand, AI teach) vs. average slop.
Almost all hires through network, many ex-customers: from Instacart, Webflow, Shopify quit to join—had conviction, loved platform. Fun way: they're bought in, excited, see vision. Like dating—build amazing traction so best people want you. Can't beg.
Co-founder joined after seeing working early product demo.
Having fun at Gumloop—no forced late stays, flexible hours, all excited about mission. Hire filter: want to spend all time with them? 24/7 hangout test. Compounded to well-adjusted, fun, ambitious, intelligent group—momentum builds.
Million reasons not to build: obsess over moat, why X/Y company won't do it? Those people never build, become pawns. Try, prove others wrong—end up leading. Could've listed 100 reasons Zapier, OpenAI crush us—then no Gumloop.
Biggest founder quality: thinking they can do it. Blind confidence.
carry_over: ""
Keep EO in your library
Save the videos and channels worth coming back to, and find them again in one place.
